Professional Documents
Culture Documents
Conturing Spike Nozzles and Determining The Optimal Direction of The Prim Flow
Conturing Spike Nozzles and Determining The Optimal Direction of The Prim Flow
Original Russian Text © A.N. Kraiko, N.I. Tillyayeva, 2007, published in Izvestiya Rossiiskoi Akademii Nauk, Mekhanika Zhidkosti i Gaza, 2007,
Vol. 42, No. 2, pp. 194–203.
Abstract—Problems related with the optimal contouring of two-dimensional and axisymmetric spike
nozzles providing maximum thrust for given dimensions and external pressure are studied. The nozzles
under consideration are self-adjustable which is ensured by the non-zero inclination of their primary
supersonic flow to the plane (axis) of symmetry. Along with the optimal contouring of the spike, the
optimal orientation of the “primary” nozzle producing the primary flow is obtained. In the exact for-
mulation, its optimal inclination is determined by an exhaustive search for the configurations providing
maximum thrust for a given spike length and various fixed inclinations of the primary nozzle. The spike
and primary nozzle contours of these configurations are generally joined through a bend with the forma-
tion of an expansion fan in the flow around the bend. The efficiency of a simpler approximate method for
contouring the spike and determining a near-optimal inclination of the primary nozzle is demonstrated.
The method is based on passage to a modified formulation of the problem differing from the original
one in that it is only the part of the spike extending beyond the primary nozzle edge that is preassigned
rather than its entire length. In the modified formulation, there is no a bend in the flow around which an
expansion fan could be formed, while the inclination of the primary nozzle is determined in the process
of designing the unique optimal configuration.
DOI: 10.1134/S0015462807020172
Keywords: optimal contouring, spike nozzle, optimal direction of primary supersonic flows.
The optimal contouring of self-adjustable plug or spike nozzles has attracted a wide circle of researchers
(see [1–13] and book [14]). The optimal configurations with no bend in the spike contour at its junction
with the contour of the lower wall of the primary nozzle constructed within the framework of the complete
system of inviscid flow equations [1] and in an approximate formulation in [2, 3] constitute only a narrow
class of optimal spike generators realized in special cases. Generally, the contoured and given sections of
the spike are joined by a bend or a region with maximum permissible (in absolute magnitude) curvature.
These configurations were first considered in [4, 5] and later in papers [6–13]. However, both the primary
nozzle inclination and its expansion ratio were chosen from considerations related with the nozzle operating
conditions during launch, etc., rather than from the requirements for providing maximum thrust in the design
regime.
In solving the same problem, the authors of [15] returned to the formulation of the problem [1] with a
smooth spike. The primary nozzle inclination was chosen from the requirement that one of the optimality
conditions (the so-called Busemann condition) be satisfied. The discussion with the authors of [15] that
followed and the calculations carried out in the process of that discussion indicated certain important and,
at first glance, unexpected features. First, it turned out that in the case of long spikes this approach makes
it possible to obtain a near-optimal inclination of the primary nozzle and a near-optimal smooth (without
an initial bend) spike generator. As for the optimal generator, it has an initial bend, but it is small. The
results of constructing optimal self-adjustable two-dimensional spike nozzles calculating their parameters
presented below make the above-mentioned features clearer.
321
322 KRAIKO, TILLYAYEVA
Fig. 1. Diagrams of the solution of the original (a) and modified (b) and (c) problems of optimal contouring of spike
nozzles.
that is, the endface bg forming the end of curve abg, we have x = Xab for 0 ≤ y ≤ yb . As in the method
of an indeterminate control contour [14], curve x = x(y), which specifies the interval ab intersecting the
supersonic flow, is unknown beforehand. The parameter distributions along ab and the ordinate yb of the
upper part of the possible endface (for yg = 0) are also unknown.
The optimal distributions of the supersonic flow parameters along ab and the dimensions of the endface
(ordinate yb ), as well as the optimal value of ϑ0 , are determined from the solution of the variational problem
formulated below. In this problem, the unknown distributions and quantities must ensure a minimum of the
functional
b
y1 + ν p+
R = yν [ρ V 2 (cos2 ϑ − x sin ϑ cos ϑ ) + p] dy − b . (1.1)
1+ν
a
Correct to a positive factor (negative thrust of the optimized device), inessential in deriving the optimality
conditions, at fixed values of the gas flow rate
b
G = yν ρ V (x sin ϑ − cos ϑ ) dy (1.2)
a
the coordinates of point a, the greatest permissible abscissa of point b (xb ≤ Xab for yb ≥ 0), and a given
base pressure p+ ≥ 0. In Eqs. (1.1) and (1.2) y is measured from the plane or axis of symmetry, the x axis
is directed along them, x = dx/dy, p, ρ , and V are the pressure, the density, and the absolute magnitude of
the velocity, respectively, and ν = 0 or 1 in the two-dimensional and axisymmetric cases, respectively. It is
assumed that at the nozzle exit the flow is isoenergetic and isentropic; for ν = 1 it is also unswirled. We will
assume further that the flow under consideration is thermodynamically equilibrium and does not involve
shock waves. Then, if h and s are the specific enthalpy and the entropy and T is the absolute temperature,
we have
dp
2h + V 2 = const, s = const, dh − = T ds = 0. (1.3)
ρ
By virtue of Eqs. (1.3), the equations of state h = h(p, s) and ρ = ρ (p, s), and the definition of the speed
of sound, the pressure and the velocity are known functions of V such that
dp dρ ρV 1 ∂ρ
= −ρ V, =− 2 , = . (1.4)
dV dV a a2 ∂p s
2. OPTIMALITY CONDITIONS
For deriving the optimality conditions we will compose an auxiliary functional I = R + λ G, where the
Lagrangian multiplier λ is as yet unknown. For any choice of finite λ and fixed curve ab with the equation
x = x(y) and for any permissible variations of the parameters on ab and the coordinates of point b, provided
they conserve the flow rate G, the increments of I and R coincide, that is, ΔR = ΔI. Using Eqs. (1.1), (1.2),
and (1.4), we obtain
b
ΔR = ΔI = yν ρ (AV δ V + Aϑ δ ϑ ) dy + yνb (Ayb Δyb + Axb Δxb ), (2.1)
a
Here, Δyb = yb − yb and Δxb = xb − xb are the increments of the coordinates of point b for permissible
spike variations (such that they do not increase the length). Their meaning is made clear from Fig. 1c, in
which the thin lines are the C+ -characteristics of the expansion fan and the bold continuous (broken) lines
proceeding from point b are the original (varied) generators of the spike. The increments ΔR = ΔI are
determined as the differences of the mass and momentum fluxes across the broken line a f b , where a f is a
segment of the initial C+ -characteristic of the expansion fan, and curve ab. In Eq. (2.1) the coefficients Ayb
and Axb are independent of the shape of curve ab, that is, of xb .
If the unknown parameter distributions along ab and the coordinates of the point b are optimal, that is,
realize a minimum R due to the negative direction of the thrust, then for any their permissible variations the
increments ΔR = ΔI must be nonnegative.
In Eq. (2.1) the coefficients of the variations of V and ϑ on ab and the increments of the coordinates yb
and xb are functions of V , ϑ , the Lagrangian multiplier λ , the function x(y), and—in the axisymmetric case
(ν = 1)—the coordinate y. Due to the arbitrariness in the choice of x(y) we equate the coefficient AV to
zero. This results in the equation
for determining the curve ab. True, it is not known beforehand whether this curve can connect the given
point a with an as yet indeterminate point b which either lies on the vertical line x = Xab or ends to the left
of that line, in both cases satisfying certain additional conditions (see below).
We will use the arbitrariness in the choice of the Lagrangian multiplier λ in order to make the coefficient
Aϑ vanish at a certain, arbitrarily chosen (“compensating”) point k on the curve ab. Thus, at point k we let
Solving Eqs. (2.2) and (2.3) for x and λ , we find that at point k
cos(ϑ ∓ α )
x = cot(ϑ ∓ α ), = λ, (2.4)
cos α
where α is the Mach angle: sin α = 1/M. Thus, at the compensating point (and for the moment only at that
point) the curve ab is tangent to either a C− or a C+ -characteristic.
After x and λ have been chosen using the above approach, expression (2.1) for ΔR takes the form:
b
ΔR = yν ρ Aϑ δ ϑ dy + yνb (Ayb Δyb + Axb Δxb ). (2.5)
a
In Eq. (2.5) δ ϑ on the curve ab, Δyb , and Δxb can be considered to be independent. The constancy of
the gas flow rate is ensured by the fact that simultaneously with the fixed-sign variation of ϑ in the vicinity
of an arbitrary point on the curve ab different from point k or with variation of the coordinates yb and xb , ϑ
is also varied with fixed sign in a small vicinity of point k. The latter variation can always be so performed
that a given flow rate is ensured without making in the vicinity of point k a first-order contribution to ΔR
from the quantity of fixed sign δ ϑk owing to the fulfillment of the equality Aϑk = 0. By virtue of this fact
and the arbitrariness in the choice of δ ϑ in the vicinity of any point on the curve ab different from point k,
the necessary condition of optimality reduces to Eq. (2.3) everywhere on ab. As a consequence, Eqs. (2.4)
must also be fulfilled everywhere on ab, that is, ab is either a C+ or a C− -characteristic. As applied to
the problem under consideration related with the contouring of the spike of an annular nozzle or its two-
dimensional counterpart, ab is a C− -characteristic on which the optimality condition is fulfilled
cos(ϑ + α )
= λ. (2.6)
cos α
FLUID DYNAMICS Vol. 42 No. 2 2007
CONTOURING SPIKE NOZZLES 325
In accordance with [14], from Eq. (2.6) and the compatibility condition it follows that we have one more
integral on the C− -characteristic ab
a simpler, though approximate technique for determining the optimal value of ϑ0 in the original problem.
We will show how this is done and then compare the ϑ0 thus obtained with the values determined from the
complete solution of the original problem.
Thus, let the dependence
M0 = M0 (ϑ0 ) (2.10)
be known and through M0 the dependence of all the other flow parameters on ϑ0 . Let ϑa and the other
quantities with the subscript a be the flow parameters at point a on the C− -characteristic ab of the expansion
fan. In view of Eq. (2.10) and the compatibility condition for the C+ -characteristic after degeneration to
point a
cot α
dϑ + dp = 0 (2.11)
ρV 2
all the flow parameters at point a depend only on ϑ0 and ϑa . If ϑa is nonzero, then, in accordance with the
above, the optimal xb = Xab , while the optimal yb is either zero for the inequality sign in condition (2.8)
or is determined by this condition taken with the equality sign. In both cases a given gas flow rate G must
pass across the optimal characteristic determined for the chosen ϑ0 and ϑa by the equations valid on the
C− -characteristics and either of conditions (2.6) and (2.7) by integrating with respect to x from x = xa = 0
to x = xb = Xab . Therefore, there are two conditions for determining two unknowns ϑ0 and ϑa , namely, the
condition of a given gas flow rate and either yb = 0 or Eq. (2.8). The case ϑ ≡ ϑb is treated analogously.
The solution of the modified problem is particularly simple for a two-dimensional configuration all of
whose parameters are constant along ab. They are also constant on all the C− -characteristics of the expan-
sion fan located to the left of ab, in particular, on its initial characteristic ab◦ on which ϑ ≡ ϑ0 . This is
consistent with a uniform flow, inclined at the angle ϑ0 to the x axis, at the primary nozzle exit; in principle,
this eliminates the nozzle contouring problems discussed above. The same uniform flow is conserved within
the entire triangle ab◦ a◦ . Thus, in view of the constancy of the flow parameters along ab the conditions de-
termining optimal ϑ0 and ϑa in the case yb > 0 reduce to Eq. (2.10) and the equalities
pa
cot α G (ρ V sin α )a
ϑa = ϑ0 − d p, = ,
ρV 2 Xab cos(ϑ − α )a (2.12)
p0
ϑ0◦
Fig. 2. X-dependence of the optimal angle of primary nozzle inclination ϑ0 for ν = 0 and κ = 1.165 (a) and 1.4 (b).
as dependence (2.10) and to perform exact and approximate calculations over wide ranges of X and Xab ,
related by Eq. (2.13), for actual values of M0 , for example, those corresponding to the conditions of launch
from the Earth. Greater differences should be expected for smaller X and fairly high M0 , when the values
of the first and second (negative) terms on the right side of Eq. (2.13) are comparable. If X − Xab X , the
difference between the original and modified problems becomes unimportant and the difference between the
exact and approximate results insignificant.
The results of calculating the two-dimensional spike configurations (ν = 0) for the case of perfect-gas
flow with two adiabatic exponents κ = 1.165 and 1.4 are presented in Figs. 2 to 4. The points connected
by continuous and broken curves relate to the results of the calculations for the original and modified prob-
lems. The total spike length is divided by the height of the minimum section of an ideal two-dimensional
symmetric primary nozzle l ◦ with the flow rate coefficient μ = 1 and a uniform flow in the exit section.
The dimensionless flow rate and thrust were obtained by normalizing their dimensional values on ρ∗◦V∗◦ l ◦
and ρ∗◦ (V∗◦ )2 /l ◦ , respectively. Here, the degree signs refer to dimensionless quantities and the asterisks to
the critical values of the density and velocity. Therefore, G = 1, while the thrust R and the specific thrust
Rsp = R/G coincide. The base pressure p+ was taken to be zero.
In Fig. 2 the X -dependence of the optimal inclination of the primary nozzle ϑ0 (X ≥ 5) is presented
for κ = 1.165 (a) and 1.4 (b) and M0 = 1.0, 1.5, 2.0, and 2.5. In both cases, the difference between the
continuous and broken curves decreases rapidly with increase in the spike length. Figure 3 presents the
X -dependence of the difference ΔRsp (in percent; not to be confused with the increment ΔR of Section 2)
between the specific thrusts of the configurations determined from the solutions of the original and modified
problems and the thrust of an annular nozzle with an axial inlet. Since the length Xab preassigned in the
modified problem was determined from the total spike length X by formula (2.13), the optimal thrust is
greater in the original than in the modified problem. However, in accordance with Fig. 3, the excess is
substantial only for X < 10. For X ≥ 10 it is negligible even for M0 = 2.5. The reason for this is obvious
from Fig. 4 which presents the dependence of ΔRsp on the primary nozzle inclination angle obtained from
the solution of the original problem for different ϑ0 . Here, 1 is the exact thrust maximum and 2 is that for
the configuration with no spike contour bend at the initial point, that is, the solution of the modified problem
with the subsequent determination of X from Eq. (2.13). For X = 20 at M0 = 2.5 and a comparatively large
primary nozzle exit section, using a spike contour with no initial bend leads to an additional thrust loss of
0.14% for κ = 1.165 (for κ = 1.4 it is only 0.03%).
If ϑ0 = ±π /2, M0 is near-unity, p+ = p+ = 0, and the fixed length X is small, then, in accordance
with [17–20], the optimal contour of the spike (for ϑ0 = −π /2) or the supersonic part of the plug nozzle
(for ϑ0 = π /2) adjoining the given contour of the subsonic part without a bend, begins from the boundary
extremum region, where the Mach number is constant. For spike nozzles an analogous situation arises for
ΔRsp , % ΔRsp , %
Fig. 3. X-dependence of ΔRsp for ν = 0 and κ = 1.165 (a) and 1.4 (b).
ΔRsp , %
ϑ0◦
Fig. 4. ϑ0 -dependence of ΔRsp for ν = 0, M0 = 2.5, and κ = 1.165.
ϑ0 close to −π /2. On the other hand, fairly extensive calculations aimed at determining the optimal values
of ϑ0 showed that, even where such configurations are permissible, the optimal angles ϑ0 are associated
with spike contours with a bend at the initial point of the type represented in Fig. 1a. With increase in X the
magnitude of the initial bend rapidly decreases.
A natural question arises concerning the difference in the conditions determining the optimal angle ϑ0
of primary flow orientation in the original and modified problems. In the case presented in Fig. 1b the
coefficients AV and Aϑ0 on ab are zero for the optimal configuration. Therefore, at fixed G and M0 varying
the angle ϑ0 disturbs the flow parameters on the C− -characteristic ab, thus changing R by an amount of the
order of (Δϑ0 )2 . Therefore, dR/d ϑ0 = 0, that is, the primary flow direction satisfies the extremum condition.
In the modified problem, for the optimal spike (Fig. 1a) the same coefficients are zero on the interval hb
rather than on the entire C− -characteristic ab, so that the condition determining the optimal angle ϑ0 takes
the form:
h h
dR ΔR
≈ = 0, ΔR = ΔI = y Φ(V, ϑ , λ ) dy − yν Φ(V, ϑ , λ ) dy,
ν
d ϑ0 Δϑ0
a a
λ − V cos ϑ
Φ(V, ϑ , λ ) = ρ V sin α + p.
sin(ϑ − α )
Here, a (a ) is a point on the edge of the optimal (deflected by an angle Δϑ0 ) primary nozzle, the in-
tegrals are calculated along the intervals a h and ah of the C− -characteristic, and point h of the optimal
C− -characteristic is fixed. Using this condition is no simpler than seeking the optimal direction of the
primary flow by making a simple exhaustive search of angles ϑ0 , as described above.
Summary. The study performed makes it possible to evaluate the accuracy of the approximate method for
determining the optimal angle of inclination of the primary nozzles of self-adjustable two-dimensional and
axisymmetric configurations with a spike, as well as the effect on the nozzle thrust of replacing the optimal
spike contours with a bend at the junction with the lower wall of the primary nozzle by near-optimal smooth
contours. With increase in the spike length, the error of the approximate technique tends rapidly to zero.
The study was carried out with the support of the Russian Foundation for Basic Research (project
No. 05-01-00846) and the State Program for the Support of Leading Science Schools (project NSh-2124-
2003.1).
REFERENCES
1. G.V.R. Rao, “Spike Nozzle Contour for Optimum Thrust,” Planet. Space Sci. 4, 92–101 (1961).
2. H. Greer, “Rapid Method for Plug Nozzle Design,” ARS J. 31, 560 (1961).
3. G. Angelino, “Approximate Method for Plug Nozzle Design,” AIAA J. 2, 1834 (1964).
4. V.M. Borisov, “System of Bodies with Minimum Wave Drag,” Inzh. Zh., 5, 1028 (1965).
5. A.L. Gonor and A.N. Kraiko, “Some results of the study of optimal shapes at supersonic and hypersonic veloci-
ties,” in: A. Miele (ed.) Theory of Optimal Aerodynamic Shapes [in Russian], Mir, Moscow (1969), 455–492.
6. R.P. Humphreys, H.D. Thompson, J.D. Hoffman, “Design of Maximum Thrust Spike Nozzles for Fixed Inlet
Geometry,” AIAA J. 9, 1581 (1971).
7. J.J. Korte, A.O. Salas, H.J. Dunn, N.M. Alexandrov, W.W. Follett, G.E. Orient, and A.H. Hadid, “Multidisciplinary
Approach to Linear Aerospike Nozzle Design,” NASA Techn. Mem. No. 110326 (1997).
8. T. Rommel, G. Hageman, C.-A. Schley, G. Krulle, D. Manski, “Plug Nozzle Flowfield Analysis,” J. Propuls.
Power 13, 629 (1997).
9. G. Hagemann, H. Immich, T.V. Nguyen, and G.E. Dumnov, “Advanced Rocket Nozzles,” J. Propuls. Power 14,
620 (1998).
10. S.V. Baftalovskii, A.N. Kraiko, and N.I. Tillyayeva, “Optimal design of self-controlled spike nozzles and their
thrust determination at start,” in: AIAA 9th Intern. Space Planes and Hypersonic Systems and Technologies Conf.
Norfolk, Va, USA, 1999. AIAA Techn. Paper, No. 4955 (1999).
11. A.N. Kraiko, N.I. Tillyayeva, and S.V. Baftalovskii, “Optimal Design of Plug Nozzles and Their Thrust Determi-
nation at Start,” J. Propuls. Power 17, 1347 (2001).
12. S.V. Baftalovskii, A.N. Kraiko, and N.I. Tillyayeva, “Contouring self-adjustable spike nozzles optimal when
operating in a vacuum and determining their thrust at launch from the Earth,” in: Proc. 22nd Readings on Cosmo-
nautics [in Russian], Voina i Mir, Moscow (1999), 93.
13. J.J. Korte, A.O. Salas, H.J. Dunn, N.M. Alexandrov, W.W. Follett, G.E. Orient, and A.H. Hadid, “Multidisciplinary
Approach to Linear Aerospike Nozzle Design,” J. Propuls. Power 17, 93 (2001).
14. A.N. Kraiko, Variational Problems of Gasdynamics [in Russian], Nauka, Moscow (1979).
15. A.S. Kiselev, V.G. Lushchik, L.E. Sternin, and A.E. Yakubenko, “Loss of Specific Impulse Due to Friction in
Spike Nozzles,” Fluid Dynamics 36, 743 (2001).
16. A.N. Kraiko and V.V. Shelomovskii, “On Profiling Planar and Axisymmetric Nozzles and Channels to Achieve a
Given Supersonic Stream at the Exit Section,” Fluid Dynamics 16, 94 (1981).
17. A.N. Kraiko, A.S. Telyakovskii, and N.I. Tillyayeva, “Profiling an Optimal Contour of a Supersonic Nozzle with
Considerable Flow Deflection,” Zh. Vychisl. Mat. Mat. Fiz. 34, 1444 (1994).
18. A.N. Kraiko and N.I. Tillyayeva, “Optimal Profiling of the Supersonic Part of a Plug Nozzle Contour,” Fluid
Dynamics 35, 945 (2000).
19. A.N. Kraiko, K.S. P’yankov, and N.I. Tillyayeva, “Profiling the Supersonic Part of a Plug Nozzle with a Nonuni-
form Transonic Flow,” Fluid Dynamics 37, 637 (2002).
20. E.V. Myshenkov, E.V. Myshenkova, and N.I. Tillyayeva, “Numerical Investigation of the Flows in Cumulative
Short-Plug Nozzles within the Framework of the Reynolds Analogy,” Fluid Dynamics 38, 482 (2003).